Stability and Robustness of the DMC Algorithm as Compared with the Traditional ARMA Model

Abstract

A traditional linear, unconstrained optimal predictive control algorithm is considered. As models are used the Dynamic Matrix (DM) model and the Auto-Regressive Moving Average (ARMA) model. A critical examination is carried out of the sensitivity in control performance and stability to changes in move suppression factor, model parameters, dead-time, and control and prediction horizons. Furthermore, the accuracy of a parameter estimation scheme for the DM model is related to the accuracy required in the parameters for a satisfactory control performance and stability margin. The algorithms developed are used to evaluate the robustness and quality of predictive optimal control schemes, where the importance of accurate estimation of dead-times is stressed.